AI-Assisted X-ray Fracture Detection in Residency Training: Evaluation in Pediatric and Adult Trauma Patients

• DOI: 10.3390/diagnostics14060596
• Publication Date: 2024-03-12T07:55:34Z
• Authors: Mathias Meetschen; Luca Salhöfer; Nikolas Beck; Lennard Kroll; Christoph David Z...; Benedikt Michael ...; Michael Forsting; Shamoun Mizan; Lale Umutlu; René Hosch; Felix Nensa; Johannes Haubold
• Keywords: ddc:610; Medizinische Fakultät » Universitätsklinikum Essen » Institut für KI in der Medizin (IKIM); Medicine (General); diagnostic imaging; Medizin; 610; X-rays -- fractures -- bone -- artificial intelligence -- diagnostic imaging -- quality improvement; fractures; artificial intelligence; bone; Article; quality improvement; 03 medical and health sciences; R5-920; 0302 clinical medicine; Medizinische Fakultät » Universitätsklinikum Essen » Institut für Diagnostische und Interventionelle Radiologie und Neuroradiologie; 616; X-rays; ddc:610; ScholarlyArticle

Background: This study aimed to evaluate the impact of an AI-assisted fracture detection program on radiology residents’ performance in pediatric and adult trauma patients assess its implications for residency training. Methods: study, conducted retrospectively, included 200 radiographs from participants aged 1 95 years (mean age: 40.7 ± 24.5 years), encompassing various body regions. Among these, 50% (100/200) displayed at least one fracture, totaling hundred thirty-five fractures, assessed by four residents with different experience levels. A machine learning algorithm was employed detection, ground truth established consensus among two experienced senior radiologists. Fracture accuracy, reporting time, confidence were evaluated without AI support. Results: Radiology sensitivity improved significantly support (58% vs. 77% AI, p < 0.001), while specificity showed minor improvements (77% 79% = 0.0653). stand-alone achieved a 93% 77%. reduced interpretation time average approximately 2.6 s (p 0.0156) increased resident findings 0.0013). Conclusion: enhanced residents, particularly benefiting less It does not compromise reduces contributing efficiency. underscores AI’s potential radiology, emphasizing role training improvement.